\nThis course is aimed at statisticians who are new to or with limited experience of machin e learning. \;
\n
\nWhat is the benefit of attending ? \;
\n\n
Attendees will learn about a range of topics in machine learning\, including practical sessions in R.
\nOvervie w
\nFour sessions will include ML foundation (including an introduc tion\, data exploration for ML and dimensionality reduction and feature se lection)\, Supervised learning (including support vector machines and mode l evaluation and interpretation)\, model optimization and unsupervised lea rning (including clustering) and advanced topics (including neural network s\, deep learning and large language models).
\nCost
\n
\n
PSI Members: \;£\;360 +VAT
\n
\nEarly
Bird Non-PSI Members: \;£\;430 +VAT
\nNon-P
SI Members: \;£\;470 +VAT
\n*Please note: Non-M
ember rates include PSI membership until 31 Dec. 2026.
Regis tration
\nEarly Bird registration closes on Friday 5th Sept
ember.
\nTo register for this event\, please \;<
span style="text-decoration: underline\;">click here.
Agenda
\n\nDay 1: ML Foundation
\n- \n
- Definitions and terminology \n
- Key aspects of ML workflow \n
- Planning \n
- Pre-processing \n
- Modelling\, cross validation\, hyperpar ameter estimation \n
- Evaluation \n
- Reporting \n
- &ldquo\;ML in a day&rdquo\;. Overview of simple ML workflow using R packages dplyr\, caret\, ggplot2 using logistic regres sion. This will be used as a starting point to exemplify key aspects of ML \, and as a point of comparison for further methods such as Elastic Net (r egularised regression)\, gradient boosting and neural networks. \n
- Common pitfalls and key considerations for using ML models in practice \, and reviewing ML analyses\, with a focus on the analysis of clinical tr ial data \n
- \n
Day 2: Supervised learning
\n- \n
- What is supervised learning? \n
- Definitions and examples of scenarios \n
- Supervi sed learning workflow (short recap) \n
- Supervised Algorithms:\n
- \n
- LASSO and Ela stic Net \n
- Decision trees and Random Forest \n
- Support Vector Machines \n
- Evaluation and interp retation of these algorithms \n
- Practical session in R: Applica tion of above methods in caret \n
Day 3: Unsupervised learning
\n- \n
- Defi nitions and comparison to supervised learning \n
- K-means\, PCA\ , Non-linear approaches such as tSNE \n
- Worked examples in R\n
Day 4: Neural Networks and Deep Learning
\n- \n
- DEMO: ML Using Caret in R \n
- Preparing d ata: data splitting and preprocessing \n
- Building and compa ring different ML algorithms \n
- Introduction to neur al networks and basic architectures \n
- Deep learning\, Large La nguage models\, Computer Vision \n
- Discussion of Machine Learni ng vs Statistics: what are the differences\, how much to they overlap\, wh en to use what \; \n
- \n
Speaker details
\n| \n Speaker< /p>\n | \n \n Biography \n | \n
\n A bstract \n | \n
| \n <
img src="https://www.psiweb.org/images/default-source/default-album/202405
13-jolyon-faria-5043ea0dcbff3ad665b3a176ff00001f6b97.jpg?sfvrsn=2fdbaedb_0
&\;sf_site_temp=true&\;sf_site=00000000-0000-0000-0000-000000000000&
amp\;MaxWidth=200&\;MaxHeight=200&\;ScaleUp=false&\;Quality=High&
amp\;Method=ResizeFitToAreaArguments&\;Signature=7ACD722FC760A28E9EE5B3
87C8234604" data-method="ResizeFitToAreaArguments" data-customsizemethodpr
operties="{'MaxWidth':'200'\,'MaxHeight':'200'\,'ScaleUp':false\,'Quality'
:'High'}" data-displaymode="Custom" alt="20240513-Jolyon Faria-5043" title
="20240513-Jolyon Faria-5043" />Jolyon Faria | \n \n Jolyon Faria is in the Clinical Data Sciences L ung Team within Oncology Data Science &\; AI\, Oncology R&\;D\, in A straZeneca. His role is to lead and perform retrospective analyses of AZ c linical trials\, Real World Data and multimodal data\, with a focus on Sta tistics and Machine Learning\, to inform the AZ Strategy for instance for PhIII Investment Decisions. He has a background in Biology: PhD (Univ. Lee ds)\; Postdoc (Princeton Univ.)\, and Applied Statistics: (MSc Univ. Oxfor d) and is a Chartered Statistician (CSTAT\; Royal Statistical Society\, UK ). \n | \n \n | \n
 \n Moira Verbelen \n Computational Methodology Lead\, UCB \n \n
| \n \n Mo ira Verbelen is Computational Methodology Lead in the Advanced Methods and Data Science team at UCB and is passionate about adopting and implementin g cutting-edge AI methods in pharmaceutical research\, in particular deliv ering AI solutions throughout all phases of clinical development. With a b ackground bridging Pharmaceutical Sciences and Biostatistics\, Moira holds a PhD from King&rsquo\;s College London\, specializing in Machine Learnin g applied to Pharmacogenetics \n | \n \n This session introduces supervised learning in t he context of pharmaceutical data analysis\, focusing on its definition an d practical applications. We'll briefly discuss the supervised learning wo rkflow before diving into key algorithms including LASSO\, Elastic Net\, D ecision Trees\, Random Forests\, and Support Vector Machines. Emphasis wil l be placed on how to evaluate and interpret these models effectively. Par ticipants will gain hands-on experience using the caret package in R. \ n | \n
\n  \n \n Leo \;Souliotis | \n \n Leonidas Souliotis is par t of the CDS Hematology team within the Oncology Data Science at AstraZene ca. His role includes developing statistical and ML methods and tools for evidence synthesis in the Heme space. He is also leading an internal cours e on how to use Python for Data Analysis. He is a Data Camp instructor\, a fter developing the &ldquo\;Efficient Python coding in pandas". He has a b ackground in Statistics: MSc (Imperial College London)\, PhD (University o f Warwick). \n | \n \n
\;\n \; \n | \n
\n  \n \n Dan de Vassimon Manela | \n \n Dan de Vassimon Manela is a Machine Learning R&\;D and Innovation Lead at UCB and is interested in using Probabilistic Machine Learning methods to solve clinical problems across the pharmaceutical development chain. He has a b ackground in Physics (BA\, MSc Cambridge) and Statistical Machine Learning (MSci UCL). \n | \n \n
\;\n \; \n | \n
END:VEVENT END:VCALENDAR